Different energy dispersive detectors are available for spectroscopic measurements of X-ray spectra, such as proportional counters (PC), PIN-diodes and silicon drift detectors (SDD). The physical principles of these detectors are very different. A PC uses a mixture of a noble gas with a quench gas that is ionized by an incident radiation. The electrons are accelerated in the electric field and can ionize other gas atoms resulting in an internal amplification. In this case the energy for generation of the primary signal is relatively high.

In solid state detectors (PIN, SDD), the incident radiation generates charge carriers which are collected by an electrical field. The energy to generate the primary signal is significantly lower than for a PC, i.e. a higher number of primary charge carriers can be generated. This reduces the statistical error and improves the energy resolution.

This report studies the analysis performance of the M1 MISTRAL when equipped with a PC or an SDD and compares it with theoretical values of a PIN-diode.

Instrumentation

All measurements were performed using the M1 MISTRAL spectrometer equipped either with a large area proportional counter or with a high resolution silicon drift detector (Bruker‘s XFlash® 5030) . The M1 MISTRAL features following technical parameters:

Analysis

Energy resolution

The energy resolution of a PC for Mn-Ka radiation is on the order of 900 eV, for PIN-diodes about 190 eV and for an SDD even less than 150 eV.

The better energy resolution of solid state detectors offers following advantages:

Below image shows the acquired spectra of a jewelry alloy sample comprised primarily of Au and with contributions of Ag, Cd, Cu and Zn, measured with a PC and an SDD. The difference in energy resoultion is demonstrated by the fact that the PC‘s spectrum does not show the contribution of Zn and that a separation of Ag and Cd is not possible.

Spectra measured with PC (blue) and SDD (red) of a jewelry alloy (concentrations approx. Au: 80%, Ag: 5 %, Cd: 5 %, Zn: 6%)

The limits of detection are also different in both cases. For a spectrometer with a PC, they are on the order of 0.5 %, whereas for a SDD the limit improves to around 0.01 %, i.e. by a factor of 50.

Sensitive area

The energy resolution of a spectrometer is determined by the detector and the electronics. The electronic noise depends mainly on the detector capacity. Therefore, detectors with better energy resolution need to be smaller, which results in a smaller sensitive area and poorer radiation collection. On the other hand, due to the smaller dimensions, the solid state detectors can be positioned closer to the sample. This compensates partly for the smaller sensitive area.

To get comparable count rates, the collimator size can be adjusted. In a PC, the spot size goes down to 0.3 mm or even smaller, but for PIN or SDD it should not be smaller than 0.5 mm.

Count rate capability

The PIN-diode and the SDD feature very different count rate capabilities. At similar input count rates, the SDD delivers a significantly higher amount of net counts (see table 1). The detectors‘ dead times for an input count rate of approx. 8 kcps are also shown in table 1. The dead time is relatively high in the case of PIN-diodes, whereas it is typically very small for SDD.

Comparison of detector properties

Regarding prop-counters, peak shifts can occur due to high charge concentration close to the counting wire in cases of detection of high count rates or high energetic X-rays. The low charge carrier mobility influences the collection time and enhances the dead time.

The M1 spectrometer has a special stabilization system for the PC that allows count rates up to 50 kcps.

Stability

The stability of the detectors used in the M1 MISTRAL is similar due to Bruker‘s good stabilization electronics. Both short and long term stability are mainly determined by the counting statistics, what results in better stability at higher count rates.

The below image, which illustrates a long term repeated measurement with an SDD. The sample was measured several times over a period of approx. 120 hours for 60 s each time. It can be seen that the measured intensity is statistically distributed.

Diagram showing the normalized intensity over the time.

The behaviour of a PC is very similar, although the statistical error is even smaller due to the larger sensitive area. The calculated standard deviations for both measurements are shown in table 2. These results lead to the following statements:

Results of repeated measurements

Accuracy

The accuracy of analysis depends mainly on the intensity error. This error is determined by the collected intensity i.e. by the statistical error, by the error for the peak fitting procedure and by the spectrometer stability.

The PC‘s statistical error is smaller. On the other hand, the error resulting from peak fitting is smaller in case of less overlapping. In this case, the system is also more robust against small peak shifts.

Accuracy determination results

Diagram showing the accuracy of a prop-counter and an SDD for a measurement time of 60 s.

In order to test the accuracy, around 80 jewelry reference samples were measured with both type of detectors. The concentration of Au varied from approx. 35 % to 100 %. Figure 3 shows the deviation from the given value for these measurements. The results are summerized in table 3. This shows that the disadvantegeous counting statistics of the SDD is compensated by a smaller error in peak deconvolution.

Investment

The investment costs are different for every type of detector. PIN-diodes are more expensive than PC by a factor of 3 - 5. An SDD equipment is in turn the high end solution.

Conclusions

The detector election should be based on the type of analytical task the user is usually confronted with. For the analysis of unknown samples with flexible qualitative composition and for the detection of small concentrations, solid state detectors are the best choice. However, prop-counters are well suited for quality control with known qualitative composition.

Both detector types are very similar in terms of accuracy. The smaller statistical error of the PC is compensated by the SDD‘s fewer overlapping errors. The detectors‘ performance regarding stability is also alike.